Patterns of care for prostate cancer patients in the United States have changed as early detection has improved. The National Cancer Data Base of the American College of Surgeons Commission on Cancer provides information about the treatment of patients in all age, race, and regional groups from institutions that represent cancer care at the community level as well as in medical centers.
Data on 251,416 prostate cancer patients diagnosed between 1992 and 1994 were studied. Patient and disease characteristics, including age, race, the geographic region from which a case was reported, American Joint Committee on Cancer stage, and tumor grade were related to the primary pattern of treatment. Stage-standardized comparisons were made among different age groups, race groups, and regional groups.
Stage and tumor grade varied little across age, race, and regional classifications. African American patients were more likely than white patients to have Stage IV prostate cancer at the time of diagnosis. Men older than 75 years had greater proportions than younger men of both the earliest and the most advanced stages of cancer. Overall, 24.6% of patients received no cancer-directed therapy, 11.6% were treated with hormones or endocrine surgery, 28.6% received radiation therapy, 28.3% underwent radical prostatectomy, and 6.9% were treated by other modalities or combinations of modalities. Treatment patterns varied markedly by age. The selection of radical prostatectomy relative to no treatment and radiation therapy varied by race and region.
The number of prostate cancer patients treated annually in the U.S. has increased as prostate specific antigen (PSA) and other early detection measures have been widely implemented.1, 2 Reports from specialized data bases, reimbursement claims data, and individual institution series have documented trends in selected aspects of care.3-6 However, these important and informative sources do not depict the broader spectrum of prostate cancer care as it occurs in patients of all ages in all regions of the country. Herein, the authors describe variations in prostate cancer treatment in the U.S. as evident from the most recent contents of the American College of Surgeons Commission on Cancer National Cancer Data Base (NCDB).
MATERIALS AND METHODS
The NCDB is a hospital-based information resource on patterns of cancer patient care and treatment outcomes. Comparisons of the NCDB with population-based registry data have demonstrated that the NCDB is representative of cancer care at the community level in the U.S.7 NCDB data are collected using a computerized, standard format that has been described elsewhere.8 Hospital participation is voluntary and hospital cancer committees or their equivalent oversee data collection and quality assurance.
A total of 259,026 prostate cancer patients diagnosed in the years 1992, 1993, and 1994 were reported to the NCDB. Ninety-eight percent of patients were age 50-89 years at diagnosis, younger and older patients were excluded from these analyses. After further exclusion of 1% of reports for which the regional source could not be accurately determined, a total of 251,416 cases were available for analysis. The number of hospitals contributing reports for 1992 was 1144, was 996 for 1993, and was 1188 for 1994.
American Joint Committee on Cancer (AJCC) Stage 0 (T1aN0M0) includes impalpable, well differentiated tumors for which cancer was an incidental histologic finding for <5% of resected tissue with no regional lymph node metastasis (N0) and no distant metastasis (M0). Stage I (T1a-c) includes more poorly differentiated impalpable tumors and tumors identified by needle biopsy resulting from elevated PSA. Stage II (T2a-c) includes tumors palpable or visible on imaging that are confined within the prostate. Stage III tumors (T3a-c) involve extension through the prostatic capsule and/or invasion of the seminal vesicles. Stage IV (T4) includes tumors that are fixed or invade adjacent structures other than the seminal vesicles and/or involves lymph nodes (N1-3) or distant sites (M1).9
Tumor registrars often report pathologic stage to the exclusion of clinical stage data. The stage used in the current study was based on the clinical stage, which was supplanted by pathologic stage when it was reported. A consequence of this procedure is that the stage depends to some extent on treatment. It is possible, for example, that a Stage II cancer treated by radiation therapy would be classified as Stage III had it been treated by radical prostatectomy. This could occur as a result of the extent of disease information obtained from surgery that is unavailable from radiation treatment.
Tumor grade was reported as well differentiated (Grade 1), moderately differentiated (Grade 2), poorly differentiated (Grade 3), or undifferentiated or anaplastic (Grade 4). Although Gleason's grading is used conventionally for prostate cancer, it is not reported to the NCDB, which relies on a common data set for all tumor types.
Treatments included all therapy performed or planned at the time of diagnosis regardless of where the treatment was administered. Surgical procedures were coded following the Registry Operations and Data Standards manual.10 Radical prostatectomy was defined as a total prostatectomy (excised prostate, ejaculatory ducts, and seminal vesicles) with or without lymph node dissection. Subtotal prostatectomy or more extensive surgery such as cystoprostatectomy or pelvic exenteration was classified as "Other Treatment." Transurethral resection of the prostate was not considered cancer-directed treatment, and patients who received only this type of treatment were included in the "No Treatment" category. Radiation therapy included external beam and interstitial treatment. Hormone treatment included medical treatment and endocrine surgery (i.e., bilateral orchiectomy). Patients who were reported to have received combinations of these modalities were classified as "Other Treatment." Patients who received chemotherapy, biologic response modifiers, or experimental treatment also were included in "Other Treatment." Any patient for whom no radical prostatectomy, radiation therapy, hormone treatment, or other treatment was reported was classified as "No Treatment," thereby exhausting all possible patterns of care.
"White" race included whites of Spanish and non-Spanish origin, "African American" included all patients reported as black, and "Other" includes patients of Asian descent, American Indians, and other or unknown race classifications.
The definitions of regions are as follows: Northeast: Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, and Vermont; Southeast: Delaware, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, Washington, DC, and West Virginia; Midwest: Illinois, Iowa, Indiana, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin; South: Alabama, Arkansas, Kentucky, Louisiana, Mississippi, Oklahoma, Tennessee, and Texas; Mountain-Pacific: Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming, Alaska, California, Hawaii, Oregon, and Washington.
This report is descriptive. No hypotheses are subjected to formal statistical analysis. The numbers of patients included in the comparisons are very large. Statistical tests for differences in proportions have little value because small differences that are unimportant from a public health or clinical perspective nevertheless are statistically significant.
Stage, Race, Age, and Region
Overall, 3.5% of cases were AJCC Stage 0, 19.2% were Stage I, 40.7% were Stage II, 13.3% were Stage III, 10.3% were Stage IV, and 12.9% were of unknown stage. The pattern of stage of disease varied by patient, age, race, and regional source. The data in Table 1 shows that African American patients were more likely to be classified as having metastatic disease (Stage IV) compared with white patients (16.6% vs. 9.6%). The greatest differences in stage of disease according to patient's age were observed for the oldest patients, in whom there were greater proportions of both the least advanced (Stage 0 and Stage I) and most advanced (Stage IV) tumors. The distribution of stage of disease was relatively uniform across other age categories. The proportion of patients reported as stage unknown was greater for cases reported from the Mountain-Pacific region. There was a corresponding deficit in the Mountain-Pacific region reports of Stage I patients. The stage distribution across other regions was relatively uniform.
Table 1. Distribution of Stage of Disease Overall and by Race, Age, and Region
Treatment by Grade
Tumor grade had little association with patient age or race (data not shown). African American patients had a greater proportion of poorly differentiated tumors compared with whites (22.0% vs. 18.8%). Conversely, 46.8% of African American patients had moderately differentiated tumors compared with 52.0% of whites. Tumor grade was evenly distributed across regions with the exception of a lower proportion of patients in the Mountain-Pacific region with well differentiated tumors compared with those from all other regions combined (15.7% vs. 19.1%). More Mountain-Pacific region patients also had moderately differentiated tumors compared with patients of all other regions (55.7% vs. 50.4%). Table 2 shows the association of tumor grade with treatment. Patients with well differentiated prostate cancer were most likely to fall into the "No Treatment" category. Patients with moderately differentiated tumors were most likely to be treated with radical prostatectomy.
Table 2. Distribution of Prostate Cancer Treatments by Tumor Grade
Treatment by Stage
Table 3 reports the patterns of treatment according to stage of disease. Radical prostatectomy and radiation therapy were the treatments for nearly identical proportions of patients, with 56.9% of all patients receiving one of these primary treatment modalities. Nearly one-quarter of patients (24.6%) were reported to have received no cancer-directed treatment. Hormone treatment was given to nearly half of patients (48.6%) with metastatic prostate cancer (Stage IV) but very infrequently for patients with any less advanced disease (Stages 0-III). Conversely, "No Treatment" was the most typical pattern of care for patients with earliest stages of prostate cancer. Radiation therapy and radical prostatectomy were reported most often for patients with Stage II or Stage III disease.
Table 3. Distribution of Prostate Cancer Treatments by Stage of Disease
Treatment by Age
Table 4 relates treatment patterns to patient age. These data were adjusted for the potential confounding effect of stage differences between age groups. Cell frequencies were weighted so that the patients in each age group had the same distribution of stage of disease as the patients of all other age groups. These data show that treatment selection varied markedly by patient age. Older patients infrequently are treated by radical prostatectomy. Only 5.5% of patients age > 75 years underwent radical prostatectomy compared with 48.0% of patients age 50-64 years. In contrast, younger patients less often received no treatment or hormone treatment. Radiation therapy was used at twice the frequency in patients age > 75 years compared with patients age 50-64 years (34.7% vs. 17.4%). Hormone treatment was > 3 times more likely in patients age > 75 years compared with patients age 50-64 years (21.0% vs. 6.4%).
Table 4. Stage-Standardized Distribution of Prostate Cancer Treatments by Age of Patient in Percent
Treatment by Region
Stage-standardized rates of treatment by the different modalities were calculated according to region. The "Hormone" and "Other Treatment" categories varied in a narrow range across regions (data not shown). Hormone treatment was used in a low of 10.9% of patients in both the Northeast and Southeast and in a high of 12.4% of patients in the South. The "Other Treatment" category was reported least often in the South (5.8%) and most often in the Mountain-Pacific region (7.6%). More marked variation across regions was observed for the "No Treatment," "Radical Prostatectomy," and "Radiation Therapy" categories. Figure 1 (9K) shows that "No Treatment" was reported most often from the Northeast Region and least often from the Mountain-Pacific region (28.0% vs. 22.2%). A similar ranking was observed for "Radiation Therapy," with 33.1% of all patients in the Northeast region undergoing radiation therapy compared with only 24.0% in the Mountain-Pacific region. Conversely, radical prostatectomy was chosen by a greater proportion of patients in the Mountain-Pacific region than by patients of comparable stage in the Northeast region (34.4% vs. 21.4%).
Treatment by Race
Stage-standardized rates for the "Hormone" or "Other Treatment" categories varied little across race groups (data not shown). Eleven percent of white patients received hormone treatment compared with 13.7% of African American patients and 14.0% of patients of other and unknown race. Seven percent of white patients were in the "Other Treatment" category compared with 6.4% for African American patients and 6.0% of patients of other or unknown race. Greater variation occurred in the distribution of the "No Treatment," "Radical Prostatectomy," and "Radiation Therapy" categories by patient race (Fig. 2 (7K)). African American and white patients were equally likely to be treated by radiation therapy. However, when compared with African American patients, White patients were more likely to undergo radical prostatectomy (29.2% vs. 22.3%) and less likely to elect to receive no treatment (23.9% vs. 29.1%).
These results may describe the norms of prostate cancer in the U.S. for the time period they cover. The years encompassed by these data may be particularly significant in representing the culmination of a period of rapid change in the incidence of prostate cancer in the U.S. After the demonstration of the effectiveness of PSA testing combined with digital rectal examination, transrectal ultrasound, and fine-needle biopsy, prostate cancer incidence increased very markedly. This effect, which began approximately 5 years earlier, reached its maximum impact in 1993. Incidence rates have declined modestly since 1993.11 Data that describe patterns of care in earlier years are based on a population of tumors detected by different means and having different potentials for effective treatment. Thus, these data may be particularly significant as benchmark descriptions of prostate cancer care at the peak of incidence after the introduction of PSA testing in well men.
The data suggest that factors other than tumor characteristics influence treatment patterns for prostate cancer. Stage of disease and tumor grade varied relatively little across race, age, or region classification but treatment patterns did. This was evident even after the effects of differences in extent of disease were controlled by stage-standardization.
Other investigators have reported unique patterns of care for elderly prostate cancer patients.12 Age may be relevant to treatment selection because life expectancy is an important factor in evaluating the risks and potential benefits of treatment as well as the option of no treatment. Age also potentially relates to different levels of concern patients may have about the risk of treatment complications such as impotency.
In contrast, the basis for the correlation between region and treatment choice is not self-evident. The lower frequency of well differentiated tumors in the Mountain-Pacific region might account for the lower proportions of patients in the "No Treatment" category. However, it is unclear why that region would have fewer well differentiated tumors. The more frequent selection of radical prostatectomy in the Mountain-Pacific region compared with the Northeast may not be wholly attributable to either differences in the characteristics of patients or the tumors detected. Whether these variations reflect differences in the training of physicians or other aspects of the medical care environment can not be determined from these data alone. It also is possible that this effect is a manifestation of variations in patient preferences across region. Although regional variation in prostate cancer treatment has been suggested in earlier reports,3, 4 little research has been done to assess the underlying reasons for its occurrence.
Race-related variations in prostate cancer treatment patterns also are poorly understood. Racial or socioeconomic discrimination might contribute to African American patients being less likely to undergo radical prostatectomy and to more often opt for no treatment. These variations have been reported by other investigators.13 However, if discrimination was the underlying factor affecting treatment variation, reduced availability of radiation therapy or hormone treatment for African American patients also might be expected. Although racial and socioeconomic discrimination may play some role, the fact that only the greater selection of observation over radical prostatectomy appears to be affected may indicate that variations reflect processes other than discrimination. This is another deserving topic that has been the subject of little systematic inquiry.
The data used in these analyses have the strengths of having been obtained from every region of the U.S. from institutions that deliver cancer care at the community level. The number of patients studied was very large and tumor registry personnel familiar with data quality standards collected the information in a standardized format. However, some caveats should be noted. Because of the scale of the data collection process, it is not possible to audit and verify all reports. Some uncorrected error is unavoidable. Hospitals and patients are not selected for inclusion on the basis of systematic sampling. It is known that the smallest hospitals are underrepresented in the NCDB and hospitals that do not maintain tumor registries are not represented at all. However, previous comparison of the NCDB with the National Cancer Institute population-based Surveillance, Epidemiology, and End Results program showed prostate cancer data to be nearly identical.7
An additional shortcoming in the data is the amalgam of clinical and pathologic staging information. Although perhaps less important for other cancer sites included in the NCDB, prostate cancer stage can differ significantly depending on the basis for its determination. This weakness in the data is regrettable but it is likely to affect different regions and populations equally. If this is the case, inferences from these data may not be impacted by the manner of stage reporting.
There is uncertainty concerning the optimal treatment of patients with prostate cancer. The age- race-, and region-related variations in prostate cancer treatment observed in the current study may be engendered to some extent by the absence of definitive standards for treatment selection. These variations in treatment merit continued monitoring.